Krzysztof Cyran and Marek Kimmel (2010) have presented a revised set of estimates of the human mtDNA most recent common ancestor (MRCA). It's an interesting theoretical paper, written for the purpose of developing a method that doesn't rely on the same assumptions as the usual coalescent models.
Their new method gives an estimate of 174,000 years ago for the human MRCA. They report an upper/lower range as 96,000 to 449,000 years ago. That range does not represent a confidence interval on the estimate, it's an upper/lower based on extreme assumptions about human/Neandertal genetic distance and the human/Neandertal MRCA.
The Neandertal mtDNA has really affected the way we estimate human MRCA, at least for the mitochondrial genome. Chimpanzees are just too distant. When we compare human and chimpanzee mtDNA genomes, there has been a lot of parallelism and reversal on both lineages, because mutations have hit the same place multiple times. Multiple hits and purifying selection make a mess out of rate estimation -- generally, they make the human MRCA seem a lot older than it truly was. The Neandertals are closer, and are therefore less of a problem.
But the Neandertal-human MRCA itself was poorly known, as long when we had only chimpanzees to calibrate the mutation rate.... That's what we discovered earlier this year with the mtDNA genome of the Denisova specimen
Still, the paper gives enough detail to work out the effect of a lower human-Neandertal MRCA on their estimate. They obtained their lower bound (96,000 years) by assuming a human-Neandertal MRCA of 389,000 years. If we substitute in the Denisova-informed human-Neandertal MRCA, we can figure that the human MRCA will be around 130,000 years ago or so.
That's awfully recent.
I don't want to go too far with these numbers. My first objection is that they all assume the total absence of selection, when we have long known that some human mtDNA clades have been selected in some parts of the world. It's entirely possible that the human MRCA is recent because of natural selection on some mitochondrial-linked phenotype ("Complete Neandertal mitochondrial sequence, and selection on human (not Neandertal) mtDNA", "Has the dam broken on mtDNA selection?", "Selection, nuclear genetic variation, and mtDNA").
And even if we assume no selection at all, there's not a lot to be gained by increased precision of these estimates. Branch lengths of an mtDNA genealogy give only extremely wide estimates of ancient events. Saying that something happened "around 50,000 years ago, plus or minus 35,000", it hardly matters whether we change that to "around 43,200 years ago, plus or minus 35,000." I would even argue that the round estimate is better, because it doesn't communicate a misleading impression of precision.
Still, it does a lot of good to know whether estimates are systematically biased in one direction. And this work, combined with what we know about the Neandertal and Denisova complete mtDNA genomes, suggests that our mtDNA branch lengths may have been biased too high.
It remains to be seen how much of the human mtDNA tree will be affected by this logic. The most recent branches can in many cases be calibrated against historical events, and ultimately parent-offspring comparisons. So those aren't likely to change much. What worries me is that critical period around 30,000--80,000 years ago, when human mtDNA lineages were diversifying worldwide. The timescale of mtDNA divergence is already out of whack with the rest of the genome. Pushing these divergences more recent will make the fit between mtDNA and autosomal estimates worse. But given the wide variance on coalescence times, Cyran and Kimmel's estimates are consistent with the hypothesis that these might be substantially higher -- so it's hard to guess whether the apparent mismatch is real or not.
I might have missed this paper if it weren't for the press release about it from Rice University. But what a misleading release! It's headlined, "Mother of all humans lived 200,000 years ago" -- which the paper doesn't conclude. If that were the conclusion, it wouldn't be news, because it's confirming a widely-used estimate that's more than 20 years old.
But there are actually several interesting angles to the story that the press release fails to mention. Their estimation method may prove useful for many species for which we have no good demographic model -- a problem that the release alludes to, but doesn't feature. The method they develop came from a similar process, which had formerly led to a much, much higher estimate of human MRCA. Their estimate is a lot lower -- in large part because they can exploit the Neandertal genetic information. And then there's the likely possibility that the actual MRCA may be much lower, which would truly be unexpected compared to most earlier work.
At the end of their paper, Cyran and Kimmel give a short discussion of the history of the Out of Africa mtDNA story. They mention the idea that some people favoring the multiregional hypothesis had suggested older dates for the human mtDNA MRCA. Aside from O'Connell
It's sort of interesting that even in the current paper, we still have an upper estimate of the human MRCA that's nearly 450,000 years ago! I don't think that the assumptions going into that value are realistic, but there's no real upper confidence bound on the central estimate. It might well go as high as 450,000 years, given the huge uncertainty in the depth of the deepest branches of that African mtDNA genealogy.
So I guess I'm not really sure we've advanced very far in 20 years!